Early changes in passively sensed homestay predict depression symptom improvement during digital behavioral activation.

Digital behavioral activation (BA) is scalable, accessible, and efficacious for depression. However, some individuals do not improve during digital BA, and identifying non-responders early is critical for facilitating adaptive intervention approaches (e.g., stepped care). To explore whether passive sensing data might serve as early predictors of symptom change, we tested whether early changes in passively sensed behavioral targets of BA predicted depression symptom changes during app-based BA. Young adults (N = 47) with elevated depressive symptoms completed a 12-week trial of an app-based BA intervention, Vira. The Vira app provided BA psychoeducation, assessed self-reported daily mood, and used smartphone sensors to passively assess time spent at home (i.e., homestay), walking, stationary time, time in bed, bedtime, and waketime each day. We quantified early behavioral changes by fitting a multilevel growth model for each behavior over the first 2 weeks of the intervention. Models included a random slope reflecting each participant's average day-to-day change in that behavior. We extracted these slope estimates and tested whether they predicted depressive symptom (PHQ-8) change from pre-to post-intervention. We hypothesized that individuals with greater early changes in intervention-targeted behaviors would experience greater reductions in depressive symptoms. As hypothesized, individuals with greater early decreases in passively sensed homestay (i.e., reduced behavioral withdrawal) experienced a greater reduction in depressive symptoms by the end of treatment (b = 0.94, p = .025). Early changes in other behaviors did not significantly predict depressive symptom change (ps > .158). Passively monitoring early changes in homestay during app-based BA may support the early identification of individuals at risk of symptom persistence, thus providing earlier opportunities to adjust treatment.